Full-waveform inversion (FWI) can be applied to time-lapse (4D) seismic data for subsurface reservoir monitoring. However, nonrepeatability (NR) issues can contaminate the data and cause artifacts in the estimation of 4D rock and fluid property changes. Therefore, evaluating and studying the NR effects on the 4D data and FWI results can help, for instance, discriminate inversion artifacts from true changes and guide seismic survey design and processing workflows. Using realistic reservoir models, data, and field measurements of NR, we find the effects of NR source-receiver position and seawater velocity changes on the data and the 4D FWI results. We show that ignoring these NR effects in the inversion can cause strong artifacts in the estimated velocity change models and thus should be addressed before or during inversion. The NR source-receiver positioning issue can be successfully addressed by 4D FWI, whereas the NR water velocity issue requires measurements or estimations of water velocities. Furthermore, we compare the accuracy and robustness of the parallel, double-difference, and central-difference 4D FWI methods to realistic NR ocean-bottom node data in a quantitative way. Parallel 4D FWI fails to capture geomechanical changes and also overestimates the aquifer layer changes with NR data. Double-difference 4D FWI is capable of recovering the geomechanical changes, but it is also sensitive to NR noises, generating more artifacts in the overburden. By averaging the forward and reverse bootstrap 4D estimates, central-difference 4D FWI is more robust to NR noises and also produces the most accurate 4D estimates.

中文翻译：

---

海底节点数据延时 4D 地震全波形反演的不可重复性影响

全波形反演 (FWI) 可应用于延时 (4D) 地震数据以进行地下储层监测。然而，不可重复性 (NR) 问题会污染数据并在 4D 岩石和流体属性变化的估计中造成伪影。因此，评估和研究 NR 对 4D 数据和 FWI 结果的影响可以帮助区分反演伪像和真实变化，并指导地震勘测设计和处理工作流程。使用真实的储层模型、数据和 NR 的现场测量，我们发现了 NR 源-接收器位置和海水速度变化对数据和 4D FWI 结果的影响。我们表明，在反演中忽略这些 NR 效应会在估计的速度变化模型中造成强烈的伪影，因此应该在反演之前或期间解决。NR 源接收器定位问题可以通过 4D FWI 成功解决，而 NR 水速问题需要测量或估计水速。此外，我们以定量的方式比较了并行、双差和中心差 4D FWI 方法与现实 NR 海底节点数据的准确性和鲁棒性。并行 4D FWI 无法捕捉地质力学变化，并且还高估了 NR 数据的含水层变化。双差 4D FWI 能够恢复地质力学变化，但它也对 NR 噪声敏感，在覆盖层中产生更多伪影。通过对前向和反向自举 4D 估计求平均，中心差分 4D FWI 对 NR 噪声更加稳健，并且还产生最准确的 4D 估计。